A framework to characterize, model, and optimize water-energy systems: case study of a novel HDH system

Abstract

The rapid development in water-energy systems calls for advanced modeling and optimization tools to improve the technologies throughputs and minimize their energy consumption. While having a wide footprint in multiple research areas, Artificial Intelligence applications in water-energy systems are still in early stages. There is a need for generalized models to fully characterize parametric relationships, optimize technologies' performance, and have the flexibility to apply to different complex and non-linear systems. This article describes a digital twin framework that can be used for that purpose. This framework provides fundamental understanding of the process as in process-driven models alongside with the accuracy and usability of data-driven methods. The framework's structure focused on modeling and optimization aspects of digital twin models. As a part of developing a novel humidification-dehumidification desalination technology, this framework was used to model energy consumption based on 16 different operational parameters and was accurate to within 1 % of the data. Pairwise parameters sensitivities were identified by the framework for two case studies involving the whole cycle operation and a critical subsystem of the technology. Moreover, the framework was used to optimize the technology performance through data-efficient Bayesian search methods. Different search realms were used in this process. The framework was able to identify minimum and maximum energy consumption regions by sampling only 25 data points out of possible 5157 cases.